Re: How my BBC Sounds Plugin briefly broke LMS. A cautionary tale!
Michael Herger <slim-b/[email protected]>
| Newsgroups | gmane.music.equipment.slimdevices.devel |
|---|---|
| Message-ID | <[email protected]> |
I would absolutely agree that platform support is becoming a bit long in the tooth. Not only for the many combinations of OS/CPU/Perl versions. But eg. for Windows and macOS, too! - Windows: we're using ActiveState Perl Development Kit's (PDK) PerlApp and siblings, based on ActivePerl 5.14. ActiveState recommends to NOT use the PDK any more. They actually suggest applications should be run using Docker. Which IMHO isn't an option for us on Windows (yet). - macOS has long stuck with Perl 5.18. Only in Mojave they installed 5.28. On Big Sur we have 5.30 in addition to the previous two. With Apple's switch to the M1 CPU we're additionally facing the fact that we don't have the platform support yet. I tried to build dependencies for 5.30 on M1 but failed (so far). - macOS and Windows come with some additional tools, like a tray bar, or the preference pane. In particular the macOS PrefPane is in dire need of some love: I'm still running builds in a MacOS 10.5(!) VM using the old XCode from back then, because newer version wouldn't compile any more... - Linux: much discussed dependency hell. So there's much more work than I can handle. And as long as things run... > Officially (supported) Debian systems are Stretch and Buster. Stretch comes > with perl 5.24 and Buster comes with 5.28. So it is reasonable to drop > everything older than 5.24. It's not that easy. Some platforms aren't as well supported as the Pi, and updates come in slower. Even NAS vendors tend to update slowly. People keep their NAS around for years. Yes, we can drop them, and I'm trying to get people to use a Pi instead. But for many this still is "too techy". I'm looking forward to seeing PolyVection's offering. Anyway: the good news is that nothing stops you and/or other community members from coming up with new build scripts, or package definitions or whatever they're called, to create a lean .deb package fully relying on external dependencies. It looks as if somebody had done it for NixOS (see https://github.com/Logitech/slimserver/issues/138). I'd be happy to provide such a "clean" .deb package in addition to today's offerings - it could replace the old builds in the longer term. But I can't handle it. If you want LMS to change the way you're so vocal about, then you'll have to get your hands dirty. As of today there are only a few contributors to LMS itself, and they concentrate on core functionality. > Assume LMS 10 would be a breaking new version. If the current 8.x do not > automatically update I do not foresee big problems. FWIW: it's very unlikely there will ever be a LMS 10. As with Windows 9, that number has been burnt in other ways. The "UE Media Library", supporting the UE Smart Radio, internally used that version number. But then, maybe by the time we get to v10 UEML is history, and nothing is left from it. By the current pace of major updates it won't happen in this decade anyway :-D